CN102110154B - File redundancy storage method in cluster file system - Google Patents
File redundancy storage method in cluster file system Download PDFInfo
- Publication number
- CN102110154B CN102110154B CN 201110042143 CN201110042143A CN102110154B CN 102110154 B CN102110154 B CN 102110154B CN 201110042143 CN201110042143 CN 201110042143 CN 201110042143 A CN201110042143 A CN 201110042143A CN 102110154 B CN102110154 B CN 102110154B
- Authority
- CN
- China
- Prior art keywords
- data slice
- state
- data
- file
- redundancy check
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000011084 recovery Methods 0.000 claims abstract description 25
- 238000012986 modification Methods 0.000 claims description 11
- 230000004048 modification Effects 0.000 claims description 11
- 238000012423 maintenance Methods 0.000 claims description 8
- 238000000151 deposition Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 description 16
- 238000005516 engineering process Methods 0.000 description 6
- 238000006243 chemical reaction Methods 0.000 description 2
- 239000012467 final product Substances 0.000 description 2
- 208000019901 Anxiety disease Diseases 0.000 description 1
- 230000036506 anxiety Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000006116 polymerization reaction Methods 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Abstract
The invention provides a file redundancy storage method in a cluster file system. On the premise that the reliability and the availability of data are ensured, the data can be dynamically switched in different storage states. When the storage resources of the system are insufficient, the data can be converted into a redundancy check mode from a mirroring mode and stored, so occupation of a redundancy space is reduced, and the utilization rate of the space is improved. Meanwhile, the data are written and updated in the mirroring mode, so the problem how to perform redundancy calculation and redundancy date of the redundancy check mode does not need to be considered for the write request of a client; and the newly written data block is mirrored onto redundancy storage management equipment, so compared with the redundancy check mode, the write performance is greatly improved. In addition, when a storage node fails, recovery operation can be performed in a mode of combining mirroring and redundancy check so as to reduce the calculation quantity and improve the efficiency of data recovery.
Description
Technical field
The present invention relates to the Security Assurance Mechanism of the data of shared storage system, particularly the redundant storage method of group system file.
Background technology
In the data-centered information age, how appropriate protected data effectively is one of key problem of storage system.People can stand delay unusually machine, all application programs of computing machine and restart even hardware damage, but they require information to lose never.No matter the most important task of storage system is that what's wrong, to guarantees that all canned data can not lose, and as possible high-quality data, services is provided incessantly.The breaking-up of data message and lose the business continuance that not only has influence on enterprise, even greatly threaten the existence of a mechanism.
In order to guarantee to be stored in the safety of data in the disk; Those skilled in the art has proposed raid-array (RAID) technology; This technology is with the synthetic disk array of a plurality of disk groups; And in each disk the redundant information of other disk of storage, make after certain disk in the array breaks down the data on the redundant information recovered failure disk that can store according to other disk in the array.RAID representes with RAID0-RAID7 respectively according to realizing that principle can be divided into different ranks.There is bigger difference in the mode of operation of the RAID system of different stage, and more representational is the RAID1 and the RAID5 that adopts the redundancy check mode that adopts the mirror image mode.
In network store system, can adopt the technology of similar above-mentioned RAID equally, a plurality of memory nodes are formed a network RAID system.If but adopt the mirror image mode of similar RAID1, and although readwrite performance is high, space availability ratio has only 50%, the cost performance of total system is lower on the contrary.If adopt the redundancy check mode of similar RAID5, increased space availability ratio, but the especially little write performance of write performance is lower.
How to be the cluster file system of high-performance, jumbo storage Network Based, the highly reliable and high available security mechanism of data efficiently is provided, promptly keep the good performance and the loss of minimizing redundant space, this problem has become a research focus.The parallel file system PanFS of Panasas company of current cluster storage manufacturer has realized file-level redundant storage mechanism; Small documents uses the mirror image mode of similar RAID1 to deposit; And big file adopts the redundancy check mode of similar RAID5 to deposit, but PanFS still can't solve the problem that above-mentioned traditional mirror image and redundancy check technology cann't be solved.
Summary of the invention
The objective of the invention is to overcome the defective of above-mentioned prior art; Cluster file system for high-performance, jumbo storage Network Based; Highly reliable and the high available security mechanism of data efficiently is provided, in the good performance of maintenance, reduces the loss of redundant space.
The objective of the invention is to realize through following technical scheme:
The invention provides a kind of cluster file system file redundant storage method, wherein, in said file system, file leaves network storage node in the mode of data slice, comprises following store status: state 1, and data slice is deposited with the mirror image mode; State 2, data slice is deposited with the redundancy check mode; Said method comprising the steps of:
Read each data slice of the file of the state of being in 1;
Carry out redundant computation, generate the redundancy check sheet;
The redundancy check sheet is write file;
The storage space of each mirror image data sheet of releasing document;
Modification state 1 is a state 2.
File redundant storage method in the cluster file system according to the preferred embodiment of the invention, the redundancy check mode that store status 2 is adopted is parity checking.
File redundant storage method in the cluster file system according to the preferred embodiment of the invention, the step of said method are to be carried out by the application server as client in the file system.
File redundant storage method in the cluster file system according to the preferred embodiment of the invention, the store status of file also comprises state 3, wherein the original version data slice is deposited with the redundancy check mode, writes the redaction data slice that renewal generates and deposits with the mirror image mode;
Said method is further comprising the steps of:
It is constant that the original version data slice of the state that is in 2 of renewal is write in maintenance, generates the redaction data slice;
The redaction data slice is write corresponding memory node, do mirror back-up simultaneously to another memory node;
Modification state 2 is a state 3.
File redundant storage method in the cluster file system according to the preferred embodiment of the invention also comprises the following step of updating of writing:
To the data slice of depositing with the mirror image mode, revise data slice and the mirror image thereof that leaves two memory nodes in simultaneously;
To the data slice of depositing with the redundancy check mode, the original version data slice that maintenance is deposited with the redundancy check mode is constant, generates the redaction data slice; The redaction data slice is write corresponding memory node, do mirror back-up simultaneously to another memory node.
File redundant storage method in the cluster file system according to the preferred embodiment of the invention, further comprising the steps of:
Traversal is in the mirror image data sheet of the file of state 3 successively, runs into the cavity, the original version data slice is filled write redaction data slice and mirror image data sheet;
The original version data slice is replaced with the redaction data slice;
The storage space of release new edition data sheet and redundancy check data sheet;
File redundant storage method in the cluster file system according to the preferred embodiment of the invention also comprises the recovering step when data slice is made mistakes:
During state 1, with the data slice image recovery error data sheet on the corresponding node;
During state 2, carry out the data slice that the redundant computation recovery makes mistakes;
During state 3,, then carry out redundant computation and recover to make mistakes data slice if what make mistakes is the original version data slice; If what make mistakes is redaction data slice or its mirror image data sheet, then with the data slice image recovery error data sheet on the corresponding node.
Compared with prior art, the file redundant storage method in the cluster file system that provides according to the specific embodiment of the invention, data can dynamically be switched in different storage states.For the write operation of client, data are write renewal with the mirror image mode, needn't consider how to carry out redundant computation and redundant replacement problem, the data block of newly writing is mirrored on the redundant storage management equipment for the request of writing of client.Avoided the low problem of write performance that the small letter renewal causes in the redundancy check mode, write performance is compared existing redundancy check mode and is greatly improved.Simultaneously, the non-data of enlivening character write can be converted into the redundancy check mode and store when system's storage resources is nervous, under the prerequisite that guarantees data reliability and availability, have reduced redundant space and have taken, the raising space availability ratio.When memory node breaks down, can reduce calculated amount in addition, improve the efficient that data are recovered, ensure the reliabilty and availability of data with the mode that mirror image mode and redundancy check combine.
Description of drawings
Followingly the embodiment of the invention is described further with reference to accompanying drawing, wherein:
Fig. 1 is not for writing the synoptic diagram of the original state of data according to each memory node of the embodiment of the invention;
Fig. 2 is the synoptic diagram of the state 1 deposited with the mirror image mode according to the data slice of the embodiment of the invention;
Fig. 3 is the synoptic diagram of the state 2 deposited with the redundancy check mode according to the data slice of the embodiment of the invention;
Fig. 4 is for to deposit with the redundancy check mode according to the original version data slice of the embodiment of the invention, writes the synoptic diagram of the state 3 that redaction data slice that renewal generates deposits with mirror image mode redundancy;
Fig. 5 is for writing the synoptic diagram of the sheet that Updates Information for 1 time according to the state of the embodiment of the invention;
Fig. 6 is for write the synoptic diagram after the renewal once more according to 3 times redaction data slice of state of the embodiment of the invention;
Fig. 7 is for being write the synoptic diagram of renewal according to 3 times original version data slice of state of the embodiment of the invention;
The synoptic diagram of Fig. 8 for makeing mistakes according to 2 times redundancy check sheets of state of the embodiment of the invention;
The synoptic diagram of Fig. 9 for makeing mistakes according to 2 times data slice of state of the embodiment of the invention;
The synoptic diagram of Figure 10 for makeing mistakes according to 3 times redundancy check sheets of state of the embodiment of the invention;
The synoptic diagram of Figure 11 for makeing mistakes according to 3 times original version data slice of state of the embodiment of the invention;
Figure 12 is the synoptic diagram of makeing mistakes according to the mirror image of 3 times redaction data slice of state of the embodiment of the invention;
Figure 13 is the synoptic diagram of makeing mistakes according to the data slice of 3 times original version data slice of state of the embodiment of the invention and redaction thereof.
Embodiment
In order to make the object of the invention, technical scheme and advantage are clearer, pass through specific embodiment to further explain of the present invention below in conjunction with accompanying drawing.Should be appreciated that specific embodiment described herein only in order to explanation the present invention, and be not used in qualification the present invention.
Said in an embodiment of the present invention file system comprises as application server, meta data server and a plurality of memory node, and wherein application server is as the read-write operation of the client executing file of file system.Fig. 1 is an original state according to an embodiment of the invention, and each memory node (SN) does not write data.When preserving file for the first time in an embodiment of the present invention, file leaves the corresponding network memory node respectively in the mode of data slice, simultaneously each memory node data slice is made mirror back-up to all the other nodes, forms mirror image mode redundancy and deposits, and improves data reliability.This moment, the store status of file was called state 1, and is as shown in Figure 2.Fig. 2 is the state 1 according to the embodiment of the invention, and data slice is deposited with the mirror image mode.Be example only in the present embodiment, but the number of memory node is unrestricted in other embodiments, can be arbitrary number with 3 memory nodes and 3 mirror nodes.
In an embodiment of the present invention, the store status of said file system file also comprises state as shown in Figure 32, and data slice is deposited with the redundancy check mode.What redundancy check mode in the present embodiment adopted is the parity checking of similar RAID5, and redundancy check mode in other embodiments can adopt also that other are redundancy check mode that those skilled in the art knew like Hamming code verification etc.
In an embodiment of the present invention, above-mentioned file storage state 1 can convert state 2 into through following steps, and switch process comprises:
Read each data slice of the file of the state of being in 1;
Carry out redundant computation, generate the redundancy check sheet;
The redundancy check sheet is write file;
The storage space of each mirror image data sheet of releasing document;
Modification state 1 is a state 2.
Above step also can be called degraded operation.When being in state 1, preserves heap file with the mirror image mode; The file space redundancy reaches 100%, if this moment, system's residue storage resources was comparatively nervous, can carry out degraded operation to file; State 1 is converted into state 2; Preserve file with redundancy check, on the basis that does not reduce system reliability, discharge storage resources, reduce space hold.
Degraded operation is carried out by the application server as client in the file system in embodiments of the present invention, can avoid producing the single Control Node performance bottleneck of centralized management like this.This degraded operation can be called by the backstage finger daemon (DegradeDeamon) of demoting, and when resource is nervous, carries out space compression and resource release.In certain embodiments, degraded operation also can be carried out by storage managing server.In further embodiments, the degradation step can be carried out by the server end in the file system.
The store status of said in embodiments of the present invention file system file also comprises state as shown in Figure 43, and wherein the original version data slice is deposited with the redundancy check mode, writes the redaction data slice that renewal generates and deposits with the mirror image mode.The store status of file is transformed into state 3 from state 2 when the file that is in state 2 being write renewal, may further comprise the steps:
It is constant that the original version data slice of the state that is in 2 of renewal is write in maintenance, generates the redaction data slice;
The redaction data slice is write corresponding memory node, do mirror back-up simultaneously to another memory node;
Modification state 2 is a state 3.
Above step also can be referred to as the write operation to state 2.As shown in Figure 4, the file that is in state 2 is write renewal, to write the sheet D that Updates Information
10Be example, keep original old film section D
10Constant, and generate new segment D
11, simultaneously with new segment D
11Mirror image D
11' be saved to another memory node, realize mirror back-up, to guarantee the reliability of data, this moment, the store status of file was called state 3.
In an embodiment of the present invention file is write when upgrading and need be carried out different operation according to the different location modes of data slice, may further comprise the steps:
To the data slice of depositing with the mirror image mode, revise data slice and the mirror image thereof that leaves two memory nodes in simultaneously;
To the data slice of depositing with the redundancy check mode, the original version data slice that maintenance is deposited with the redundancy check mode is constant, generates the redaction data slice; The redaction data slice is write corresponding memory node, do mirror back-up simultaneously to another memory node.
Above step can be referred to as to write upgrades operation.Fig. 5 writes the synoptic diagram of the sheet that Updates Information for 1 time for state in embodiments of the present invention; During state 1; Data slice is deposited with the mirror image mode; When it is write renewal, need revise the data slice and the mirror image thereof that are kept at two memory nodes with the mirror image mode simultaneously, modification can be adopted synchronous mode or asynchronous mode.To write the sheet D that Updates Information
1Be example, upgrade D
1Be D
12, simultaneously with the mirror image D of another memory node
1' be updated to D
12'.This moment, state data memory still was a state 1, and above step also can be called upgrades operation to writing of state 1.
The file that is in state 3 is write when upgrading, distinguished the storage mode of data slice, the original version data slice is deposited with the redundancy check mode, deposits with the mirror image mode and write the redaction data slice that renewal generates.Fig. 6 is for being write the synoptic diagram after the renewal according to 3 times redaction data slice of state of the embodiment of the invention, wherein the redaction data slice is deposited with the mirror image mode, and the file of the state that is in 3 shown in Figure 4 is write renewal, upgrades D to write
11Be example, keep the original original version data slice D that deposits with the redundancy check mode
10Constant, upgrade D
11Be D
12, simultaneously with D
11Mirror image D at another memory node
11' be updated to D
12', realize mirror back-up, this moment, store status also was a state 3.
Fig. 7 is for being write the synoptic diagram of renewal according to 3 times original version data slice of state of the embodiment of the invention; Wherein the original version data slice is deposited with the redundancy check mode, and the file that is in state 3 as shown in Figure 6 is write renewal, upgrades D to write
30Be example, keep original version data slice D
30Constant, and generate redaction data slice D
31, simultaneously with redaction data slice D
31Mirror image D
31' be saved to another memory node, realize mirror back-up, to guarantee the reliability of data.This moment, store status also was a state 3.
Read-write operation
In embodiments of the present invention, application server is write read operation as client.
Client is carried out read operation, can not influence the file system coherency state, can not influence the establishment and the maintenance of snapshot, therefore get final product by normal flow process, and like step (1), (2), shown in (3):
(1) client is to service end demand file layout relationship information layout (being that the file logical address is to physical address map);
(2) service end inquiry associated metadata returns corresponding layout;
(3) client is read the data on the memory device through the layout that returns.
Application server carries out write operation as client, like step (4), (5), shown in (6):
(4) client is to service end request layout information;
(5) service end inquiry associated metadata returns corresponding read-write layout, and reserves corresponding resource.Client is read and write according to the equipment that carries out of the layout that returns, and according to the content that read-only layout reads memory device, the content of modification writes can be write in the layout corresponding stored equipment;
(6) client is submitted corresponding metadata information to layout, and joins reserved resource information in the metadata tissue.
For read request, after client obtained the mapping resources address through meta data server, from the direct request msg of memory node, performance was not influenced by redundant storage; Needn't consider redundant replacement problem for writing requesting client, the data block that only needs to use the Remote Switched Port Analyzer mode newly to write is mirrored on the redundant storage management equipment, and write performance is compared the redundancy check mode and is greatly improved.Client also can adopt and postpone to write and write the performance that technology such as polymerization and buffer memory are further optimized write operation in certain embodiments.
The store status of file can be transformed into state 1 from state 3 through following steps in an embodiment of the present invention:
Traversal is in the mirror image data sheet of the file of state 3 successively, runs into the cavity, the original version data slice is filled write redaction data slice and mirror image data sheet;
The original version data slice is replaced with the redaction data slice;
The storage space of release new edition data sheet and redundancy check data sheet;
Above step also can be called updating operation.Operated as follows as client by application server in the present embodiment: traversal is in the mirror image data sheet block ' of the file of state 3 successively, runs into the cavity, explains that then this data slice does not write valid data as yet, and original version data slice block is described
0Be not modified, do not generate redaction data slice block naturally yet
iFill the original version data slice and write redaction data slice block this moment
iWith mirror image data sheet block ';
With original version data slice block
0Replace with redaction data slice block
i, because need file store status 3 of living in be converted into state 1, retention data sheet and mirror image thereof this moment;
With new edition data slice block
i, redundancy check data sheet P relevant information delete, empty and reclaim; Revising state 3 simultaneously is 1;
Successively the file that is in state 3 is carried out updating operation and just can accomplish from state 3 to state 1 conversion.
In the present embodiment, above-mentioned degraded operation, to the write operation of state 2, writing renewal operation and updating operation all is to be carried out by the application server as client in the file system.In some other embodiment, aforesaid operations also can be carried out by the server in storage managing server or the file system.
A kind of method of file storage state dynamic translation is provided for cluster file system in some embodiments of the invention, and the file in the wherein said file system comprises following store status:
State 1, data slice is deposited with the mirror image mode;
State 2, data slice is deposited with the redundancy check mode;
Described dynamic converting method may further comprise the steps:
State 1 gets the hang of 1 through writing to upgrade to operate;
State 1 gets the hang of 2 through degraded operation;
State 2 gets the hang of 3 through the write operation to state 2;
When state was 1, the data slice that degraded operation will be preserved with the mirror image mode originally changes into the redundancy check mode preserved, and is transformed into state 2.When this operation usually occurred in system memory space resource anxiety, compressible data discharged storage space.
State 2 is through the write operation to state 2; Be transformed into state 3; Keep original data slice constant at state 3 with the redundant fashion storage; Need write and directly generate the redaction data slice when upgrading operation and it is done mirror image preserve, all like this writing upgraded operation and all needn't be carried out redundant computation and upgrade with redundant, thereby improved write performance.
When the store status of file is 3,, whole or most data slice upgrades operation, then whole or most data slice D if having carried out to write
IjAll preserve mirror image D at another memory node
Ij', in fact with mirror image mode store data.Can carry out updating operation this moment.After operation was accomplished, store status returned to state 1.
Reasons such as (make mistakes like power fail, operating system etc.) or network failure (connect interrupt etc. like network) stop service because node failure when certain memory node (SN), cause loss of data after, need be with faulty equipment reparation or replacement.After accomplishing reparation or replacement, client need adopt recovery operation to recover obliterated data, guarantees the reliability of data., data slice adopts following steps in an embodiment of the present invention when makeing mistakes:
During state 1, with the data slice image recovery error data sheet on the corresponding node;
During state 2, carry out the data slice that the redundant computation recovery makes mistakes;
During state 3,, then carry out redundant computation and recover to make mistakes data slice if what make mistakes is the original version data slice; If what make mistakes is redaction data slice or its mirror image data sheet, then with the data slice image recovery error data sheet on the corresponding node.
Above-mentioned steps also can be described as recovery operation.With reference to accompanying drawing 8,9,10,11,12,13 pairs of recovery operations specify below.
In embodiments of the present invention, when file is in state 1,, possibly have only 1 data burst to small documents; As far as big file, have the plurality of data sheet and be distributed in different nodes, each data slice all has mirror back-up at another node.When file is in state 1, carrying out recovery operation only need get final product the error data sheet of malfunctioning node with the image recovery on the corresponding node.
In embodiments of the present invention, 2 times error data sheet classification of state and recovery operation are following:
As shown in Figure 8, state 2 times, the recovery operation when redundancy check sheet P makes mistakes need be read his data slice D of file
1, D
2And D
3, make redundant computation, generate new redundancy check sheet P.As shown in Figure 9, state 2 times, data slice D
3Recovery operation when makeing mistakes need be read other data slice D in the redundancy group
1, D
2And redundancy check sheet P, make redundant computation, generate new data sheet D
3
In embodiments of the present invention, 3 times error data classification of state and recovery operation are following:
Shown in Figure 10, state 3 times, redundancy check sheet P
0Recovery operation when makeing mistakes need be read other original version data slice D that deposits with the redundancy check mode
10, D
20And D
30, make redundant computation, recover redundancy check sheet P
0Shown in Figure 11, state 3 times, original version data slice D
30Recovery operation when makeing mistakes need be read other original version data slice D
10, D
20And redundancy check sheet P, make redundant computation, restore data sheet D
30Shown in Figure 12, state 3 times, mirror image data sheet D
11' recovery operation when makeing mistakes is through utilizing the data slice D of the node of not makeing mistakes
11, the mirror image data sheet D of the node that recovers to make mistakes
11'.Shown in Figure 13, state 3 times, original version data slice D
10With redaction data slice D
11Recovery operation when makeing mistakes was divided into for two steps: at first need read in other original version data slice that the file on the node of not makeing mistakes is deposited with the redundancy check mode, make redundant computation, recover original version data slice D
10Then through data slice D in the mirror image preservation
11', the data slice D on the node that recovers to make mistakes
11
Upgrade with redundant it is thus clear that the mode that when memory node breaks down, adopts mirror image and redundancy check to combine also can reduce the redundant computation amount, improve the efficient that data are recovered.
Above-mentioned in embodiments of the present invention recovery operation is to be carried out by the application server as client in file system.In some other embodiment, aforesaid operations also can be carried out by the server in storage managing server or the file system.
In some embodiments of the invention, the store status of the file in the said file system can be: state 1, and data slice is deposited with the RAID1 mode; State 2, data slice is deposited with RAID5 pattern redundancy; State 3, the original version data slice is deposited with RAID5 pattern redundant fashion, writes the redaction data slice that renewal generates and deposits with the RAID1 mode.
Specific embodiment through the invention described above; Cluster file system for storage Network Based; The method of highly reliable efficiently and high available security mechanism is provided, and the data redundancy mechanism of utilizing mirror image and redundancy check to combine guarantees the data reliability and the availability of file system.The file redundant storage method that provides according to the embodiment of the invention makes the store status of data carry out dynamic conversion: for the write operation of client; Data are write renewal with the mirror image mode; Needn't consider how to carry out redundant computation and the redundant replacement problem in the redundancy check mode for the request of writing of client; The data block of newly writing is mirrored on the redundant storage management equipment; Avoided upgrading the low problem of write performance that causes like small letter in the redundancy check modes such as RAID5, write performance is greatly improved.Simultaneously, the non-data of enlivening character write can be converted into the redundancy check mode and store when system's storage resources is nervous, under the prerequisite that guarantees data reliability and availability, have reduced redundant space and have taken, the raising space availability ratio.
It is thus clear that the file redundant storage method according to the embodiment of the invention possesses good space availability ratio, and can effectively reduce the redundant computation expense, improve system performance.When memory node breaks down, can reduce calculated amount simultaneously, improve the efficient that data are recovered, ensure the reliabilty and availability of data with the mode that mirror image and redundancy check combine.
Though the present invention is described through preferred embodiment, yet the present invention is not limited to described embodiment here, also comprises various changes and the variation done without departing from the present invention.
Claims (6)
1. cluster file system file redundant storage method, said file system file leaves network storage node in the mode of data slice, comprises following store status:
State 1, data slice is deposited with the mirror image mode;
State 2, data slice is deposited with the redundancy check mode;
State 3, wherein the original version data slice is deposited with the redundancy check mode, writes the redaction data slice that renewal generates and deposits with the mirror image mode;
Said method comprising the steps of:
Read each data slice of the file of the state of being in 1;
Carry out redundant computation, generate the redundancy check sheet;
The redundancy check sheet is write file;
The storage space of each mirror image data sheet of releasing document;
Modification state 1 is a state 2;
It is constant that the original version data slice of the state that is in 2 of renewal is write in maintenance, generates the redaction data slice;
The redaction data slice is write corresponding memory node, do mirror back-up simultaneously to another memory node;
Modification state 2 is a state 3.
2. the file redundant storage method in the cluster file system according to claim 1 is characterized in that: said redundancy check mode is parity checking.
3. cluster file system file redundant storage method according to claim 1 is characterized in that: said method is carried out by the application server as client in the file system.
4. cluster file system file redundant storage method according to claim 1 is characterized in that: also comprise the following step of updating of writing:
To the data slice of depositing with the mirror image mode, revise data slice and the mirror image thereof that leaves two memory nodes in simultaneously;
To the data slice of depositing with the redundancy check mode, the original version data slice that maintenance is deposited with the redundancy check mode is constant, generates the redaction data slice; The redaction data slice is write corresponding memory node, do mirror back-up simultaneously to another memory node.
5. cluster file system file redundant storage method according to claim 1 is characterized in that said method is further comprising the steps of:
Traversal is in the mirror image data sheet of the file of state 3 successively, runs into the cavity, the original version data slice is filled write redaction data slice and mirror image data sheet;
The original version data slice is replaced with the redaction data slice;
The storage space of release new edition data sheet and redundancy check data sheet;
Modification state 3 is 1.
6. according to claim 1,4 or 5 described cluster file system file redundant storage methods, it is characterized in that: also comprise the recovering step when data slice is made mistakes:
During state 1, with the data slice image recovery error data sheet on the corresponding node;
During state 2, carry out the data slice that the redundant computation recovery makes mistakes;
During state 3,, then carry out redundant computation and recover to make mistakes data slice if what make mistakes is the original version data slice; If what make mistakes is redaction data slice or its mirror image data sheet, then with the data slice image recovery error data sheet on the corresponding node.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110042143 CN102110154B (en) | 2011-02-21 | 2011-02-21 | File redundancy storage method in cluster file system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110042143 CN102110154B (en) | 2011-02-21 | 2011-02-21 | File redundancy storage method in cluster file system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102110154A CN102110154A (en) | 2011-06-29 |
| CN102110154B true CN102110154B (en) | 2012-12-26 |
Family
ID=44174315
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201110042143 Expired - Fee Related CN102110154B (en) | 2011-02-21 | 2011-02-21 | File redundancy storage method in cluster file system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102110154B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102750195A (en) * | 2012-06-07 | 2012-10-24 | 浪潮电子信息产业股份有限公司 | Method for cluster file system data fault tolerance |
| CN102857554B (en) * | 2012-07-26 | 2016-07-06 | 福建网龙计算机网络信息技术有限公司 | Data redundancy processing method is carried out based on distributed memory system |
| CN103761268B (en) * | 2014-01-06 | 2017-12-01 | 无锡城市云计算中心有限公司 | A kind of method of Distributed File System Data storage layout |
| CN104239417B (en) * | 2014-08-19 | 2017-06-09 | 天津南大通用数据技术股份有限公司 | Dynamic adjusting method and device after a kind of distributed data base data fragmentation |
| CN104267913B (en) * | 2014-10-20 | 2017-06-16 | 北京北亚宸星科技有限公司 | It is a kind of can dynamic asynchronous adjust RAID storage method and storage system |
| CN105550367A (en) * | 2015-06-30 | 2016-05-04 | 巫立斌 | Asynchronous remote copy method for memory |
| CN106383665B (en) * | 2016-09-05 | 2018-05-11 | 华为技术有限公司 | Date storage method and coordination memory node in data-storage system |
| CN106954098B (en) * | 2017-03-22 | 2019-04-09 | 深圳市九洲电器有限公司 | A kind of set top box upgrading data selecting method and system |
| CN108959547B (en) * | 2018-07-02 | 2022-02-18 | 上海浪潮云计算服务有限公司 | PV snapshot distributed database cluster recovery method |
| CN110764692B (en) * | 2019-08-28 | 2022-11-11 | 计算力(江苏)智能技术有限公司 | Method and system for redundancy and recovery of storage slice data in server cluster environment |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101681282A (en) * | 2006-12-06 | 2010-03-24 | 弗森多系统公司（dba弗森-艾奥） | Be used to share, front end, the device of distributed raid, system and method |
-
2011
- 2011-02-21 CN CN 201110042143 patent/CN102110154B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101681282A (en) * | 2006-12-06 | 2010-03-24 | 弗森多系统公司（dba弗森-艾奥） | Be used to share, front end, the device of distributed raid, system and method |
Non-Patent Citations (1)
| Title |
|---|
| JOHN WILKES et al..《The HP AutoRAID Hierarchical Storage System》.《ACM Transactions on Computer System》.1996,第14卷(第1期),111、119-121. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102110154A (en) | 2011-06-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102110154B (en) | File redundancy storage method in cluster file system | |
| JP6294518B2 (en) | Synchronous mirroring in non-volatile memory systems | |
| US7159150B2 (en) | Distributed storage system capable of restoring data in case of a storage failure | |
| CN102024044B (en) | Distributed file system | |
| KR100275900B1 (en) | Method for implement divideo parity spare disk in raid sub-system | |
| US11074129B2 (en) | Erasure coded data shards containing multiple data objects | |
| US20030041211A1 (en) | Dual axis RAID systems for enhanced bandwidth and reliability | |
| CN102521058A (en) | Disk data pre-migration method of RAID (Redundant Array of Independent Disks) group | |
| CN103098015A (en) | Storage system | |
| US11003558B2 (en) | Systems and methods for sequential resilvering | |
| CN103942112A (en) | Magnetic disk fault-tolerance method, device and system | |
| US7260739B2 (en) | Method, apparatus and program storage device for allowing continuous availability of data during volume set failures in a mirrored environment | |
| US20110154100A1 (en) | Apparatus and method of performing error recovering process in asymmetric clustering file system | |
| CN100543743C (en) | Multiple machine file storage system and method | |
| CN116204137B (en) | Distributed storage system, control method, device and equipment based on DPU | |
| CN105871987A (en) | High available system and method for data writing | |
| US20140173337A1 (en) | Storage apparatus, control method, and control program | |
| US8276017B2 (en) | Process, apparatus, and program for system management | |
| US20200327025A1 (en) | Methods, systems, and non-transitory computer readable media for operating a data storage system | |
| US20200363958A1 (en) | Efficient recovery of resilient spaces | |
| US20130185257A1 (en) | Cloud data resiliency system and method | |
| KR101323729B1 (en) | Apparatus and method of processing failure data recovery in the asymmetric clustering file system | |
| JP2010134788A (en) | Cluster storage device, and method of controlling same |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20121226 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |